ZOU ET AL . VOL. 6 ’ NO. 10 ’ 8997–9004 ’ 2012

www.acsnano.org

8997

September 12, 2012

C 2012 American Chemical Society

Mechanism of Polarization Fatiguein BiFeO3Xi Zou,†,^ Lu You,†,^ Weigang Chen,† Hui Ding,† Di Wu,‡ Tom Wu,§ Lang Chen,† and Junling Wang†,*

†School of Materials Science and Engineering, Nanyang Technological University, Singapore 639798, ‡Department of Materials Science andEngineering, Nanjing University, China 210093, and §Division of Physics and Applied Physics, School of Physical and Mathematical Sciences,Nanyang Technological University, Singapore 637371. ^These authors contributed equally to this work.

Fatigue of ferroelectric materials refersto the reduction of switchable polari-zation after repetitive electrical cycling.

It is detrimental to the performance andlifetime of ferroelectric-based devices, suchas ferroelectric random access memory (Fe-RAM), actuators, and microwave electroniccomponents.1 To obtain a comprehensiveunderstanding of the underlying mechan-ism, extensive studies have been devotedto fatigue in both thin film and bulk ferro-electric materials.2�5 Different models havebeen proposed, including defect redistribu-tion,6 charge injection,7 and local phasedecomposition.8 The main problem is thatthe majority of the studies focus on macro-scopic properties, for example, dielectricpermittivity and polarization-electric fieldhysteresis loop. The microscopic modelsare, in most cases, conjectures. Direct micro-scopic study of ferroelectric fatigue is scarce.Only a few reports can be found in theliterature. For example, Setter et al. showeddirect observation of a frozen domain withpreferred orientation in fatigued Pb(Zr0.45-Ti0.55)O3 film using piezoresponse forcemicroscopy (PFM).9 Gruverman et al. alsoreported a similar unswitchable polarizationdomain in Pb(Zr,Ti)O3 using different ele-ctrodes.10 Yang et al. proposed the time-dependent domain wall pinning as thefatigue mechanism in Pb(Zr,Ti)O3.

11 Thema-jority of this work performed PFM on topof the electrode or after removing the topelectrode. And sometimes it is possible todetermine the polarization configurationalong the depth profile in a vertical ferro-electric capacitor.12 However, direct imagingof space charge redistribution and domainevolution across the film thickness duringpolarization reversal is not available. Thereare also reports on the study of fatigue andpolarization switching using structure char-acterization tools. For example, Do et al.observed structural relaxation in fatigued

Pb(Zr,Ti)O3 using focused X-ray diffraction.13

Recently, Nelson et al. reported the effectsof point defect on the polarization switchingin BiFeO3 using high resolution transmissionelectron microscopy (TEM).14 But its effecton polarization fatigue is still unclear. Tocircumvent the technical hindrance from avertical sandwich structure, we developed aplanar metal�insulator�metal (MIM) struc-ture. By using a planar device, we offer threeadvantages over the vertical device. (1) Di-rect observation of domain evolution acrossthe film thickness during polarization rever-sal. (2) Mapping of space charge distribution(charged defects or injected charges) during

* Address correspondence tojlwang@ntu.edu.sg.

Received for review July 10, 2012and accepted September 12, 2012.

Published online10.1021/nn303090k

ABSTRACT

Fatigue in ferroelectric oxides has been a long lasting research topic since the development of

ferroelectric memory in the late 1980s. Over the years, different models have been proposed to

explain the fatigue phenomena. However, there is still debate on the roles of oxygen vacancies

and injected charges. The main difficulty in the study of fatigue in ferroelectric films is that the

conventional vertical sandwich structure prevents direct observation of the microscopic

evolution through the film thickness during the electric field cycling. To circumvent this

problem, we take advantage of the large in-plane polarization of BiFeO3 and conduct direct

domain and local electrical characterizations using a planar device structure. The combination

of piezoresponse force microscopy and scanning kelvin probe microscopy allows us to study the

local polarization and space charges simultaneously. It is observed that charged domain walls

are formed during the electrical cycling, but they do not cause polarization fatigue. After

prolonged cycling, injected charges appear at the electrode/film interfaces, where domains are

pinned. When the pinned domains grow across the channel, macroscopic fatigue appears. The

role of injected charges in polarization fatigue of BiFeO3 is clearly demonstrated.

KEYWORDS: multiferroic . fatigue . charged domain wall . charge injection .scanning kelvin probe microscopy

ARTIC

LE

ZOU ET AL . VOL. 6 ’ NO. 10 ’ 8997–9004 ’ 2012

www.acsnano.org

8998

fatigue. (3) Explicit distinguishing between the inter-face and bulk effect. Here we report the direct observa-tion of domain evolution and charge activities duringpolarization fatigue measurements using a combina-tion of scanning kelvin probe microscopy (SKPM)and PFM, It is observed that electron injection causesdomain pinning and eventually leads to polarizationfatigue in BiFeO3.BiFeO3 has been intensively investigated due to

the coexistence of ferroelectric and antiferromagneticorders above room temperature.15,16 The intimatecoupling between the ferroic orders makes BiFeO3 apromising candidate for magnetoelectronics.17 Recentdiscoveries of intriguing properties related to the ferro-electric domain walls further add to the functionalitiesof BiFeO3-based devices.18,19 The study of the fatiguemechanism of BiFeO3 is still in the infancy stage. Themajority of the studies focus on the improvement offatigue performance through doping, for example,La,20 Ti,21 and Gd.22 The fatigue mechanism of intrinsicBiFeO3 is still unclear. Baek et al. reported orientation-dependent fatigue behavior in single-domain BiFeO3

thin films and attributed it to local domain pinning bycharged domain walls.23 However, the domain evolu-tion and space charge distribution during fatigue ismissing. What makes BiFeO3 attractive for fatiguestudy is that it possesses a large in-plane polarizationcomponent because of the R3c structure.24 It is thuspossible to observe in-plane domain evolution andcharge activities in the films directly while fatigue isinduced using a planar device structure.25 In addition, alot of functional devices of BiFeO3 are based on planarstructure. Therefore the study of fatigue on a planarBiFeO3 device is desirable.26,27 To unravel the interac-tion between local space charges (defect-induced orinjected) and ferroelectric polarization, and to eluci-date the mechanism of fatigue in BiFeO3, we haveconducted SKPM and PFM studies on (001)-orientedBiFeO3 films using a planar device setup as shown inthe inset of Figure 1b.

RESULTS AND DISCUSSION

Stage I. Polarization Switching in a Planar BiFeO3 Device.The 40 nm-thick BiFeO3 films were deposited on (001)-oriented SrTiO3 single crystal substrates by pulsed laserdeposition (PLD). Standard photolithography and lift-off process were used to prepare the planar electrodesas shown in the inset of Figure 1b. We have studiedboth the cases when electrodes are on top of andembedded in the film, and the results are qualitativelythe same. Only results obtained from the device withelectrodes on top are reported here. The channel widthis∼9.5 μm. Before carrying out the fatigue experiment,we examined the polarization switching of the planarBiFeO3 capacitor first using the remanent hysteresismeasurement method. The in-plane remanent polar-ization of the as-deposited film is 51.9 μC 3 cm

�2, which

is in good agreement with the reported value forBiFeO3 (Figure 1a).28 The coercive field is significantlyreduced when compared with conventional verticalcapacitors, consistent with the scaling law when treat-ing the channel width as the capacitor thickness.29 Anelectric field of 100 kV 3 cm

�1 is sufficient to switch theplanar capacitor. No fatigue was observed after 1010

cycles (Figure 1b). The domain structures before andafter switching were recorded by PFM with the canti-lever aligned along the [110]pc direction of BiFeO3.Figure 1 panels e and f show the in-plane (IP) PFMimages. The color code of the IP-PFM images is shownin Figure 1c. Different from the initial crisscross domainpatterns,28 highly oriented stripe domains are formedafter electrical switching. By carefully reconstructingthe polarization directions using both OP-PFM and IPPFM images,30 we conclude that two out of the eightpossible ferroelectric variants coexist. The IP polariza-tion has the typical head-to-tail arrangement betweenadjacent domains forming 90� in-plane domain walls(insets of Figure 1e,f).

Stage II. Formation of Charged Domain Walls. The elec-trical cycling was carried out using bipolar squarepulses of constant value (100 kV 3 cm

�1) and width(0.1 ms). No fatigue was observed up to 1010 cycles(Figure 1b). However, the local domain structure dras-tically altered during the cycling. The original uniformhead-to-tail arrangement was perturbed by the emer-gence of block domains after ∼106 cycles (Figure 2a).The amount of block domains increased after furtherswitching to 108 cycles (Figure 2b,c). Between theseblock domains, either head-to-head or tail-to-tail con-figuration is identified, indicating charged domainwalls (Figure 2c). The sketches in Figure 2d show howsuch domains could have been formed. The first sketchindicates the intermediate stage of the typical 71�switching where purple stripes turn to brown andbrown stripes change to yellow. Upon reversing theelectric field, occasionally, 109� polarization switchingmay happen, that is, switching between two oppositebrown stripes (second sketch in Figure 2d) or yellowstripes turn directly to purple (third sketch in Figure 2d),leading to the block domains with charged domainwalls. We observed that the charged domain wallsare perpendicular to the electrode/film interface,consistent with the fact that polarization switchingis achieved through domain nucleation at the inter-face followed by forward domain growth across thechannel.18 What's surprising is that these chargeddomain walls do not pin the domains. By applying anopposite field, the block domains can still be switched(Figure 2e). This is consistent with the macroscopicpolarization measurement results, but contradicts theprevious claim of charged domain walls being a pos-sible cause of fatigue in vertical sandwich capacitors.10

Indeed, charged domain walls that are parallel to thepolarization switching direction should not impede the

ARTIC

LE

ZOU ET AL . VOL. 6 ’ NO. 10 ’ 8997–9004 ’ 2012

www.acsnano.org

8999

process from an electrostatic point of view, becausethe polarity and amount of bound charges at thecharged domain walls did not change during polariza-tion reversal.

The question is: Whywould such domainwalls formsince they have high electrostatic energy due to thehead-to-head or tail-to tail arrangement? In fact, thisphenomenon has been studied both experimentallyand theoretically.31,32 Charged domain walls in (Pb,Zr)TiO3 were successfully observed at atomic scale byJia et al.33 using negative spherical-aberration imagingtechniques in an aberration-corrected TEM. Brennanet al. proposed that it is caused by defect-domaininteractions.34 We suggest that charged domain wallsform occasionally during the switching process as

described above. After that, charged defects or elec-trons aggregate at the domain walls to reduce theelectrostatic energy. Since oxygen vacancy is the mostmobile defect in perovskite oxides, they will accumu-late at the tail-to-tail domain walls to compensate thenegative polarization charges. At the same time, thepositive polarization charges at the head-to-headcharged domain walls are compensated by electronsin the film. If this is the case, we should observe nospace charges (complete compensation) or residualpolarization charges (incomplete compensation) atsuch domain walls. We thus conducted SKPM imagingto test this assumption.

SKPM Characterization of Charged Domain Walls. SKPM is adual pass technique based on atomic forcemicroscope

Figure 1. Hysteresis measurement and polarization in a planar BiFeO3 device. (a) Remanent hysteresis loops of as-prepareddevice and after 1010 cycles. (b) Switchable polarization obtained from the hysteresis loops showsno fatigueup to 1010 cycles.Inset of panel b shows the planar device configuration. (c,d) Theworkingprinciple of PFMand color code of IP-PFM (c) andOP-PFM (d) images is demonstrated. The purple and yellow stripes represent IP polarization perpendicular to the AFM tip. The IPpolarization parallel to the tip is in a brown color. The OP component of polarization shows yellow and purple color forpointing up and down respectively. (e,f) PFM images obtained after applying �100 kV 3 cm

�1 and þ100 kV 3 cm�1

fielddemonstrate the IP domain switching and the head-to-tail arrangement between adjacent domains (inset of panels e and f,dash arrows represent the IP polarization components).

ARTIC

LE

ZOU ET AL . VOL. 6 ’ NO. 10 ’ 8997–9004 ’ 2012

www.acsnano.org

9000

(AFM), and is sensitive to the potential differencebetween the tip and the sample. The first scan, wherethe tip is mechanically driven at its free-standing res-onance frequency, captures the sample topography.The second scan is then conducted at a fixed distancefrom the surface with a small AC bias of frequency ωapplied to the tip. The electrostatic force between thetip and the sample drives the cantilever to vibrate atthe same frequency. Since the force is proportionalto the potential difference between the tip and thesample, it can be nullified by applying a DC bias to thetip whose magnitude equals to the original potentialdifference.35 Collecting the DC bias applied during thescan leads to the potential variation across the samplesurface. The potential obtained by the SKPM reflectsthe net space charge distribution in the sample.

In the SKPM image (Figure 3b, collected from a dif-ferent sample with the sample channel width), brightlines that run perpendicularly across the channel areobserved. These bright lines are not presented in theSKPM image for the as-deposited sample which showsuniform contrast (inset of Figure 3b). In our setup, thisindicates accumulation of net positive charges com-pared with the neighboring area,36 which can only beionized oxygen vacancies. When compared with theIP-PFM images, the bright regions in the SKPM imagematch the tail-to-tail domain walls very well as shownin Figure 3a. However, this observation suggests overcompensation at the domain walls and is differentfrom our expectation. Interestingly there is no SKPMcontrast at the head-to-head charged domain walls,indicating a complete compensation.

EFM Characterization of Charged Domain Walls. To doublecheck the nature of charges observed in SKPM, we also

conducted an electrostatic force microscopy (EFM)study. While SKPM measures a potential differencebetween the tip and sample surface, EFM measureselectrostatic forces between them directly. It is also adual pass technique, but the tip is mechanically drivenat its free-standing resonance frequency during bothscans. Following the topographic information ob-tained during the first scan, the second scan is con-ducted at a fixed distance (30 nm in our experiments)from the surface with a DC bias applied to the tip.The force between the sample and the tip alters itsresonance frequency and changes the phase andamplitude signals. In our EFM system, attractive andrepulsive forces will give rise to positive and negativephase shifts, respectively. Phase images of the BiFeO3

planar device (a different devicewith the same channelwidth) after 108 cycles switching were taken with tipbiased at�5 V andþ5 V, as shown in Figure 3 panels cand d, respectively. The perpendicular lines across thechannel coincide exactly with the bright lines in theSKPM image (Figure 3b) and the tail-to-tail domainwalls in PFM image (Figure 3a). The reversed contrastunder opposite tip biases confirms the presence ofnet positive charges (see Supporting Information fordetails, Figure S2). To explain the occurrence of over-compensation, we suggest that there are intermediatehigh energy domain walls appearing during the polar-ization switching, which drive more charges to thedomain walls. When these high energy domain wallsdisappear upon further switching, the low mobilityoxygen vacancies remain there and lead to overcom-pensation. Our preliminary study supports this model(see Supporting Information for details, Figure S3). How-ever, further investigation is needed to clarify this issue.

Figure 2. Formation of block domains with charged domain walls after electrical cycling. IP domain structure after (a) 106, (b)107, and (c) 108 cycles. Stripe domains start tomerge after 106 cycles. Charged domainwalls are clearly visible after 107 cycles.(d) Sketches of the (left) intermediate stage of normal switching process and (middle, right) the stages that lead to theformation of charged domain walls. The arrow in panel d is the direction of reversed electric field. (e) The block domainsaround the charged domain walls can still be switched when the electric field is reversed.

ARTIC

LE

ZOU ET AL . VOL. 6 ’ NO. 10 ’ 8997–9004 ’ 2012

www.acsnano.org

9001

Stage III. Electron Injection, Domain Pinning and PolarizationFatigue. Since no fatigue was observed after 108 cycleseven though charged domain walls were formed, wecontinued to apply the same bipolar electrical pulsesto the sample. After 1010 cycles, the block domainswithcharged domain walls disappeared and the domainpattern reverted back to the original stripe pattern(Figure 4a,b), and the remanent polarization remainedunchanged (Figure 1a). Furthermore, the SKPM studyshowed that the positive charges that run across thechannel also disappeared. Instead, dark regions ap-peared along the electrode/film interfaces, indicatingnegative charge accumulation (Figure 4c). Accompa-nying these changes, the device also becomes moreconductive (Figure 4d) and exhibited higher leakagecurrent. We suggest that, after prolonged electricalcycling, injected electrons compensate the originalpositive space charges. Without the defect charges,charged domain walls are not stable and the domainpattern reverts back to the original head-to-tail ar-rangement. The injected electrons also affect the inter-face barrier and possibly the resistivity of the filmaround the interface, leading to higher leakage current.More interestingly, domain pinning was clearly ob-served along the electrode/film interfaces in the IP-PFM images (outlined in Figure 4a,b). When the electricfield is reversed, the domains in these regions do notswitch. If we treat the channel width in our device as

the film thickness in a vertical capacitor, the isolateddomain pinning in the interface is consistent with theobservation by Setter and Yang et al.11,12 of frozennanodomains in the fatigued (Pb,Zr)TiO3 capacitor. Atthis stage, no reduction in the switchable polarizationwas observed macroscopically. This is because thehighly conductive dark regions in SKPM image act asextension of the Pt electrodes. Domain pinning at theinterfaces only reduces the effective channel width.Ferroelectric domains between the dark regions areswitchable whichmaintains the remanent polarization.

Upon further electrical cycling, more electrons areinjected, which continuously migrate into the film andinduce more domain pinning. We expect that whenthe pinned domains grow across the channel, macro-scopic fatiguewill be observed. Indeed, after 6.7� 1010

cycles, fatigue was observed from the macroscopichysteresis loop measurements (Figure 5a,b). Theswitchable polarization has dropped by about 44.5%from 51.9 μC 3 cm

�2 to 28.8 μC 3 cm�2. From the SKPM

image (inset of Figure 5b), both thewidth and darknessof the charge-injected areas have increased. Accom-panying the reduction in the macroscopic switchablepolarization, the IP-PFM images collected after oppo-site fields clearly demonstrate the joint of pinneddomains across the channel (red outline in Figure 5c,d).We conclude that injected electrons induce cross-channel domain pinning and cause fatigue in the

Figure 3. SKPM images of the channel with charged domain walls. (a) IP-PFM image of the planar device after 108 cycles withthe tail-to-tail charged domainwalls identified. (b) The SKPM image of the same area reveals bright lines perpendicular to theelectrode/film interface, indicating positive charges along these lines. The locations of the tail-to-tail domain walls and thepositive charge lines match very well. The bright lines were not observed in the SKPM for the as-deposited device (inset ofpanel b, in the same data scale as panel b). This indicates that the charged domainwall is due to the redistribution of chargedclusters. (c,d) The EFM images collected at the same area match SKPM and PFM images quite well at the positions of chargeddomainwalls. The inversed contrast at the charged domainwalls under�5 V (c) andþ5 V (d) tip bias indicates the presence ofnet positive charges.

ARTIC

LE

ZOU ET AL . VOL. 6 ’ NO. 10 ’ 8997–9004 ’ 2012

www.acsnano.org

9002

BiFeO3 planar device. This is different from the conclu-sions made in the study by Baek et al. where oxygenvacancies were suggested to cause the domain wallpining and lead to fatigue in BiFeO3.

23

The above discussion also suggests that the widerthe channel, the longer time/or the more electricalcycles needed to inducemacroscopic fatigue. We have

confirmed this by testing another device with a smallerchannel width of 2.5 μm. The same 100 kV 3 cm

�1

bipolar pulse was used. Indeed, upon 1.2 � 109 cycles,the dark regions around the electrode/film interfacesin the SKPM images have covered almost 2/3 of thechannel. And the IP-PFM images show joint of pinneddomains across the channel, similar to that observed in

Figure 4. Domain pinning at electrode/film interface. (a,b) Domain pinning is observed around the electrode/film interfacesafter 1010 cycles. (c) SKPM image shows negative charge accumulation around the interfaces. (d) Leakage current increasesduring fatigue. Considerable high leakage current is observed after 1010 cycles.

Figure 5. Fatigue of BiFeO3. (a) Remanent hysteresis loops indicate reduction of switchable polarization after 6.7 � 1010

cycles. (b) Switchable polarization vs cyclingnumber for the device. Inset SKPM image shows the increased injected charges atelectrode/film interfaces. (c,d) Domain pinning across the channel is demonstrated by the IP-PFM images (red outline).

ARTIC

LE

ZOU ET AL . VOL. 6 ’ NO. 10 ’ 8997–9004 ’ 2012

www.acsnano.org

9003

the 9.5μmsample. Alongwith themicroscopic changes,clear fatiguewasobserved in themacroscopic hysteresisloop (see Supporting Information for details, Figure S4).

CONCLUSIONS

It is widely accepted that ferroelectric fatigue is adefect-chemistry induced phenomena. The main-stream models proposed for the microscopic originof fatigue can be divided into two categories: redis-tribution of oxygen vacancies and charge injections. Inour planar BiFeO3 device, we observed the effects ofboth oxygen vacancies and injected charges directly.For a device with 9.5 μm channel, (1) a significantamount of charged domain walls appeared after ∼108

cycles. Surprisingly, these charged domain walls didnot impede the in-plane polarization switching underelectric field. They did not cause fatigue. (2) SKPM studyrevealed that the charged domain walls were formeddue to the interaction between local space charges,likely ionized oxygen vacancies and electrons, andbounded polarization charges during electrical cycling.

(3) The charged domain walls disappeared uponfurther switching to 1010 cycles. Meanwhile domainpinning commenced at the electrode/film interfaces.(4) SKPM study revealed negative charges accumula-tion, likely injected electrons, at the electrode/filminterfaces, where domain pinning occurred. When thepinned domains grew across the channel, macroscopicfatiguewas observed. For a devicewith 2.5 μmchannel,fatigue was observed at 1.2 � 109 cycles, which isconsiderably faster than the wider channel sample. Inconclusion, we have conducted systematic PFM andSKPMstudies on thedomain evolution of BiFeO3 duringbipolar electrical cycling using planar devices. Electroninjection at the electrode/film interfaces leads to do-main pinning. The pinned domains grow across thechannel upon further cycling and eventually lead tofatigue. The role of injected charges (electron) in ferro-electric fatigue of BiFeO3 is clearly demonstrated. Webelieve such direct observation of charge-polarizationinteraction during cycling should help to better under-stand the mechanism of ferroelectric fatigue.

MATERIALS AND METHODSMaterials. BiFeO3 films (40 nm)were grownon (001)-oriented

SrTiO3 single crystal substrate at 100 mTorr oxygen partialpressure using PLD. The substrate temperature was fixed at650 �C while the laser (248 nm) operated at 5 Hz with an energydensity of∼1 J 3 cm

�2 on the target. Pt electrodeswere preparedthrough standard lift-off process on top of BiFeO3 film with thechannel aligned along the [100]/[010] direction.

Electrical Characterization. Precision LC Ferroelectric tester(Radiant Technologies) was used to perform the remanentpolarization measurement and bipolar switching. The samplewas prepoled by a 100 kV 3 cm

�1 field for 100ms. During electricalcycling, (100 kV kV 3 cm

�1 square pulse at 0.1 ms was used.SPM Characterization. PFM and SKPM were carried out using a

commercial AFM system (Asylum Research MFP3D). Pt-coatedconductive tips (MikroMasch DPE 14, resonant frequency:160 kHz, stiffness: 5.7 n 3m

�1) were used for the scanning.During the second pass in SKPM, the tip lift height was 30 nmand an AC bias (1 V peak-to-peak amplitude) at 130 kHz wasapplied. In the second pass of EFM, the tip lift height was30 nmand vibrates at its freestanding resonance frequency. Theloading force for PFM scanning is ∼30 nN.

Conflict of Interest: The authors declare no competingfinancial interest.

Acknowledgment. The authors acknowledge the supportfrom Nanyang Technological University, Ministry of Educationof Singapore and National Research Foundation of Singaporeunder Project No. ARC 16/08 and NRF-CRP5-2009-04.

Supporting Information Available: (1) Out-of-plane domainevolution during fatigue; (2) EFM scanning of charge injection atPt/BiFeO3 interfaces. (3) hypothesis for over compensation atcharged domain walls. (4) fatigue of 2.5 μm channel planarBiFeO3 device. This material is available free of charge via theInternet at http://pubs.acs.org.

REFERENCES AND NOTES1. Scott, J. F.; Paz de Araujo, C. A. Ferroelectric Memories.

Science 1989, 246, 1400–1405.2. Ramesh, R.; Inam, A.; Chan, W. K.; Wilkens, B.; Myers, K.;

Remsching, K.; Hart, D. L.; Tarascon, J. M. Epitaxial Cuprate

Superconductor/Ferroelectric Heterostructures. Science1991, 252, 944–946.

3. Colla, E. L.; Taylor, D. V.; Tagantsev, A. K.; Setter, N.Discrimination between Bulk and Interface Scenarios forThe Suppression of the Switchable Polarization (Fatigue)in Pb(Zr,Ti)O3 Thin Films Capacitors with Pt Electrodes.Appl. Phys. Lett. 1998, 72, 2478–2480.

4. Park, B. H.; Kang, B. S.; Bu, S. D.; Noh, T. W.; Lee, J.; Jo, W.Lanthanum-Substituted Bismuth Titanate for Use in Non-volatile Memories. Nature 1999, 401, 682–684.

5. Poykko, S.; Chadi, D. J. Dipolar Defect Model for Fatigue inFerroelectric Perovskites. Phys. Rev. Lett. 1999, 83, 1231–1234.

6. Scott, J. F.; Dawber, M. Oxygen-Vacancy Ordering as aFatigue Mechanism in Perovskite Ferroelectrics. Appl.Phys. Lett. 2000, 76, 3801–3803.

7. Tagantsev, A. K.; Stolichnov, I.; Colla, E. L.; Setter, N.Polarization Fatigue in Ferroelectric Films: Basic Experi-mental Findings, Phenomenological Scenarios, andMicro-scopic Features. J. Appl. Phys. 2001, 90, 1387–1402.

8. Lou, X. J.; Zhang, M.; Redfern, S. A. T.; Scott, J. F. Fatigue as aLocal Phase Decomposition: A Switching-Induced Charge-Injection Model. Phys. Rev. B 2007, 75, 224104.

9. Colla, E. L.; Hong, S.; Taylor, D. V.; Tagantsev, A. K.; Setter, N.;No, K. Direct Observation of Region by Region Suppressionof the Switchable Polarization (Fatigue) in Pb(Zr,Ti)O3 ThinFilm Capacitors with Pt Electrodes. Appl. Phys. Lett. 1998,72, 2763–2765.

10. Gruverman, A.; Auciello, O.; Tokumoto, H. Nanoscale In-vestigation of Fatigue Effects in Pb(Zr,Ti)O3 Films. Appl.Phys. Lett. 1996, 69, 3191–3193.

11. Yang, S. M.; Kim, T. H.; Yoon, J.-G.; Noh, T. W. NanoscaleObservation of Time-Dependent Domain Wall Pinning asTheOrigin of Polarization Fatigue. Adv. Funct. Mater. 2012,22, 2310–2317.

12. Colla, E. L.; Stolichnov, I.; Bradely, P. E.; Setter, N. DirectObservation of Inversely Polarized Frozen Nanodomainsin Fatigued Ferroelectric Memory Capacitors. Appl. Phys.Lett. 2003, 82, 1604–1606.

13. Do, D.-H.; Evans, P. G.; Isaacs, E. D.; Kim, D. M.; Eom, C. B.;Dufresne, E. M. Structural Visualization of PolarizationFatigue in Epitaxial Ferroelectric Oxide Devices. Nat. Mater.2004, 3, 365–369.

ARTIC

LE

ZOU ET AL . VOL. 6 ’ NO. 10 ’ 8997–9004 ’ 2012

www.acsnano.org

9004

14. Nelson, C. T.; Gao, P.; Jokisaari, J. R.; Heikes, C.; Adamo, C.;Melville, A.; Baek, S.-H.; Folkman, C. M.; Winchester, B.; Gu,Y.; et al. Domain Dynamics during Ferroelectric Switching.Science 2011, 334, 968–971.

15. Wang, J.; Neaton, J. B.; Zheng, H.; Nagarajan, V.; Ogale, S. B.;Liu, B.; Viehland, D.; Vaithyanathan, V.; Schlom, D. G.;Waghmare, U. V.; et al. Epitaxial BiFeO3 Multiferroic ThinFilm Heterostructures. Science 2003, 299, 1719–1722.

16. Eerenstein, W.; Mathur, N. D.; Scott, J. F. Multiferroic andMagnetoelectric Materials. Nature 2006, 442, 759–765.

17. Ramesh, R.; Spaldin, N. A. Multiferroics: Progress andProspects in Thin Films. Nat. Mater. 2007, 6, 21–29.

18. Seidel, J.; Martin, L. W.; He, Q.; Zhan, Q.; Chu, Y. H.; Rother,A.; Hawkridge, M. E.; Maksymovych, P.; Yu, P.; Gajek, M.;Balke, N.; et al. Conduction at Domain Walls in OxideMultiferroics. Nat. Mater. 2009, 8, 229–234.

19. Yang, S. Y.; Seidel, J.; Byrnes, S. J.; Shafer, P.; Yang, C. H.;Rossell, M. D.; Yu, P.; Chu, Y. H.; Scott, J. F.; Ager, J. W.; et al.Above-Bandgap Voltages from Ferroelectric PhotovoltaicDevices. Nat. Nanotechnol. 2010, 5, 143–147.

20. Lee, D.; Kim, M. G.; Ryu, S.; Jang, H. M.; Lee, S. G. EpitaxiallyGrown La-Modified BiFeO3 Magnetoferroelectric ThinFilms. Appl. Phys. Lett. 2005, 86, 222903.

21. Hu, G. D.; Fan, S. H.; Yang, C. H.; Wu, W. B. Low LeakageCurrent and Enhanced Ferroelectric Properties of Ti andZn Codoped BiFeO3 Thin Film. Appl. Phys. Lett. 2008, 92,192905.

22. Hu, G. D.; Cheng, X.; Wu, W. B.; Yang, C. H. Effects of GdSubstitution on Structure and Ferroelectric Properties ofBiFeO3 Thin Films Prepared Using Metal Organic Decom-position. Appl. Phys. Lett. 2007, 91, 232909.

23. Baek, S.-H.; Folkman, C. M.; Park, J.-W.; Lee, S.; Bark, C.-W.;Tybell, T.; Eom, C.-B. The Nature of Polarization Fatigue inBiFeO3. Adv. Mater. 2011, 23, 1621–1625.

24. Li, J. F.; Wang, J. L.; Wuttig, M.; Ramesh, R.; Wang, N.; Ruette,B.; Pyatakov, A. P.; Zvezdin, A. K.; Viehland, D. DramaticallyEnhanced Polarization in (001), (101), and (111) BiFeO3

Thin Films Due to Epitiaxial-Induced Transitions. Appl.Phys. Lett. 2004, 84, 5261–5263.

25. Balke, N.; Gajek,M.; Tagantsev, A. K.; Martin, L.W.; Chu, Y. H.;Ramesh, R.; Kalinin, S. V. Direct Observation of CapacitorSwitching Using Planar Electrodes. Adv. Funct. Mater.2010, 20, 3466–3475.

26. Chu, Y.-H.; Martin, L. W.; Holcomb, M. B.; Gajek, M.; Han,S.-J.; He, Q.; Balke, N.; Yang, C.-H.; Lee, D.; Hu, W.; Zhan, Q.;Yang, P.-L.; et al. Electric-Field Control of Local Ferromag-netism Using a Magnetoelectric Multiferroic. Nat. Mater.2008, 7, 478–482.

27. Yang, C. H.; Seidel, J.; Kim, S. Y.; Rossen, P. B.; Yu, P.; Gajek,M.; Chu, Y. H.; Martin, L. W.; Holcomb, M. B.; He, Q.; et al.Electric Modulation of Conduction in Multiferroic Ca-dopedBiFeO3 Films. Nat. Mater. 2009, 8, 485–493.

28. You, L.; Liang, E.; Guo, R.; Wu, D.; Yao, K.; Chen, L.; Wang, J.Polarization Switching in Quasiplanar BiFeO3 Capacitors.Appl. Phys. Lett. 2010, 97, 062910.

29. Kay, H. F.; Dunn, J. W. Thickness Dependence of TheNucleation Field of Triglycine Sulphate. Philos. Mag.1962, 7, 2027–2034.

30. Zavaliche, F.; Yang, S. Y.; Zhao, T.; Chu, Y. H.; Cruz, M. P.;Eom, C. B.; Ramesh, R. Multiferroic BiFeO3 Films: DomainStructure and Polarization Dynamics. Phase Transit 2006,79, 991–1017.

31. Mokr�y, P.; Tagantsev, A. K.; Fousek, J. Pressure on ChargedDomainWalls and Additional Imprint Mechanism in Ferro-electrics. Phys. Rev. B 2007, 75, 094110.

32. Balke, N.; Gajek, M.; Tagantsev, A. K.; Martin, L. W.; Chu,Y.-H.; Ramesh, R.; Kalinin, S. V. Direct Observation ofCapacitor Switching Using Planar Electrodes. Adv. Funct.Mater. 2010, 20, 3466–3475.

33. Jia, C.-L.; Mi, S.-B.; Urban, K.; Vrejoiu, I.; Alexe, M.; Hesse, D.Atomic-Scale Study of Electric Dipoles Near Charged andUncharged DomainWalls in Ferroelectric Films.Nat. Mater.2008, 7, 57–61.

34. Brennan, C. Model of Ferroelectric Fatigue Due to Defect/Domain Interactions. Ferroelectrics 1993, 150, 199–208.

35. Rosenwaks, Y.; Shikler, R.; Glatzel, T.; Sadewasser, S. KelvinProbe Force Microscopy of Semiconductor SurfaceDefects. Phys. Rev. B 2004, 70, 085320.

36. Jacobs, H. O.; Leuchtmann, P.; Homan, O. J.; Stemmer, A.Resolution and Contrast in Kelvin Probe ForceMicroscopy.J. Appl. Phys. 1998, 84, 1168–1173.

ARTIC

LE